Skip to content

Benchmark

GPQA Diamond

Science benchmarks measure how well models answer science questions. Scores are percentages. Higher is better.

Models and systems measured
282
Scores recorded
1623
Sources
4
IQ weight
1.72

Score distribution

79.3%
Mean
84.5%
Median
15.7%
Std dev
56.8%
10th percentile
92.2%
90th percentile
0.168%
Min
0.168%96.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.72 for chat models, measured over 272 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or system
1051Gemini 3.1 Flash LiteOpenRouter harness, google/gemini-3.1-flash-lite-20260507, Google, public 90-day aggregate· OpenRouter benchmark exports81.3%
1052GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, River, public 90-day aggregate· OpenRouter benchmark exports81.3%
1053MiniMax M3none· Epoch AI Benchmarking Hub data81.3%
1054DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, StreamLake, public 90-day aggregate· OpenRouter benchmark exports81.3%
1055DeepSeek V4 ProOpenRouter harness, deepseek/deepseek-v4-pro-20260423, BaseTen, public 90-day aggregate· OpenRouter benchmark exports81.3%
1056Ling 3.1 FlashOpenRouter harness, inclusionai/ling-3.1-flash-20261002, Novita, public 90-day aggregate· OpenRouter benchmark exports81.3%
1057Muse Glimmer 30BOpenRouter harness, meta/muse-glimmer-30b-20260810, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81.3%
1058Qwen3.5 35B A3Bnone· Epoch AI Benchmarking Hub data81.2%
1059MiniMax M2.5OpenRouter harness, minimax/minimax-m2.5-20260211, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81.1%
1060GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Mistral, public 90-day aggregate· OpenRouter benchmark exports81.1%
1061Qwen 3.8 27BOpenRouter harness, qwen/qwen3.8-27b-20260814, ModelRun, public 90-day aggregate· OpenRouter benchmark exports81.1%
1062GPT-5 miniOpenRouter harness, openai/gpt-5-mini-2025-08-07, auto-routing, public 90-day aggregate· OpenRouter benchmark exports81%
1063Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, PrimeIntellect, public 90-day aggregate· OpenRouter benchmark exports81%
1064GPT OSS 120Bofficial model card Table 3 · reasoning high · with tools; Table 3 does not name the exact tool · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.9%
1065GLM 5OpenRouter harness, z-ai/glm-5-20260211, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports80.8%
1066OpenAI o3medium· Epoch AI Benchmarking Hub data80.8%
1067Qwen3.7 Flashnone· Epoch AI Benchmarking Hub data80.8%
1068Claude Sonnet 4.5OpenRouter harness, anthropic/claude-4.5-sonnet-20250929, Claude Platform on AWS, public 90-day aggregate· OpenRouter benchmark exports80.8%
1069Qwen3.6 35B A3BOpenRouter harness, qwen/qwen3.6-35b-a3b-20260415, DekaLLM, public 90-day aggregate· OpenRouter benchmark exports80.8%
1070GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Particle AI, public 90-day aggregate· OpenRouter benchmark exports80.7%
1071Claude Opus 4.5· Epoch AI Benchmarking Hub data80.7%
1072GPT-5 miniOpenRouter harness, openai/gpt-5-mini-2025-08-07, Azure, public 90-day aggregate· OpenRouter benchmark exports80.7%
1073DeepSeek V3.1 TerminusOpenRouter harness, deepseek/deepseek-v3.1-terminus, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports80.6%
1074GPT-5 miniOpenRouter harness, openai/gpt-5-mini-2025-08-07, OpenAI, public 90-day aggregate· OpenRouter benchmark exports80.6%
1075DeepSeek V3.2OpenRouter harness, deepseek/deepseek-v3.2-20251201, GMICloud, public 90-day aggregate· OpenRouter benchmark exports80.6%
1076LongCat-Flash-Thinking· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.80.5%
1077Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, Parasail, public 90-day aggregate· OpenRouter benchmark exports80.5%
1078Claude Sonnet 5max· Epoch AI Benchmarking Hub data80.3%
1079GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Makora, public 90-day aggregate· OpenRouter benchmark exports80.3%
1080GLM 5OpenRouter harness, z-ai/glm-5-20260211, auto-routing, public 90-day aggregate· OpenRouter benchmark exports80.2%
1081GPT OSS 120Bofficial model card Table 3 · reasoning high · no tools · pass@1· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.1%
1082Qwen 3 235B A22B Thinking 2507· Epoch AI Benchmarking Hub data80.1%
1083Qwen3.6 35B A3BOpenRouter harness, qwen/qwen3.6-35b-a3b-20260415, auto-routing, public 90-day aggregate· OpenRouter benchmark exports80%
1084Gemma 4 26B A4BOpenRouter harness, google/gemma-4-26b-a4b-it-20260403, Cloudflare, public 90-day aggregate· OpenRouter benchmark exports80%
1085DeepSeek V3.1OpenRouter harness, deepseek/deepseek-chat-v3.1, AtlasCloud, public 90-day aggregate· OpenRouter benchmark exports79.9%
1086GLM 5OpenRouter harness, z-ai/glm-5-20260211, GMICloud, public 90-day aggregate· OpenRouter benchmark exports79.8%
1087DeepSeek R1OpenRouter harness, deepseek/deepseek-r1-0528, SiliconFlow, public 90-day aggregate· OpenRouter benchmark exports79.8%
1088OpenAI o3low· Epoch AI Benchmarking Hub data79.8%
1089o4 Minihigh· Epoch AI Benchmarking Hub data79.6%
1090Claude Opus 4extended thinking · up to 64K thinking tokens· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.6%
1091Nova 2 LiteAmazon Nova 2 Lite official launch chart / technical report · accuracy, v1· Reported by Amazon Web Services, Amazon AGI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.6%
1092Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports79.5%
1093Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, Together, public 90-day aggregate· OpenRouter benchmark exports79.5%
1094GLM 5.2OpenRouter harness, z-ai/glm-5.2-20260616, GMICloud, public 90-day aggregate· OpenRouter benchmark exports79.4%
1095Kimi K2.5OpenRouter harness, moonshotai/kimi-k2.5-0127, Amazon Bedrock, public 90-day aggregate· OpenRouter benchmark exports79.3%
1096DeepSeek R1OpenRouter harness, deepseek/deepseek-r1-0528, StreamLake, public 90-day aggregate· OpenRouter benchmark exports79.3%
1097Qwen3.5 9BOpenRouter harness, qwen/qwen3.5-9b-20260310, Venice, public 90-day aggregate· OpenRouter benchmark exports79.3%
1098Muse Glimmer 30BOpenRouter harness, meta/muse-glimmer-30b-20260810, DeepInfra, public 90-day aggregate· OpenRouter benchmark exports79.3%
1099Nemotron 3 Super 120B A12B· OpenEvals Leaderboard data· Reported by the model's developer. The source relaying it did not verify it.79.2%
1100GLM 5.3 FlashOpenRouter harness, z-ai/glm-5.3-flash-20260826, Sail Research, public 90-day aggregate· OpenRouter benchmark exports79.2%

Showing 1,051–1,100 of 1,621 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or system
MiMo V2 Flash Baseofficial Base table · 5-shot· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.55.1%
Qwen3.8-2.4T-A95Blifecycle-page core-benchmark claim· Reported by Alibaba Cloud Model Studio· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.92.6%

Tokens the model generated per task, as measured by the source. Comparable within this benchmark only.

Data credits: Epoch AI Benchmarking Hub, CC BY 4.0 (licence) · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · OpenRouter public benchmark leaderboard and search-lane exports, CC BY 4.0. Fractional source scores converted to percentages; immutable dated snapshots. (licence) · OpenEvals/leaderboard-data, MIT (licence)